Evaluation of data mining for two child-related, social risk issues

Abstract

Two child-related social issues are examined using data mining to determine successful ways of predicting risk. The issues of child truancy and child abuse can be considered similar as both are influenced by, the child’s characteristics, family and environment. The results show that from an initial portfolio of algorithms, a one-nearest neighbour approach works well. We believe that reflects the nature of the problem, where expert opinion classifies each new pupil /case in terms of similar ones, while the one-nearest aspect, reflects the small amount of data we had access to.